Security is trailing adoption
Industry reporting warns enterprises are rolling out agentic workflows faster than they can secure and monitor them — creating shadow AI, permission creep, and non‑human identity gaps. Vendors are responding with agentic defensive platforms, identity‑monitoring tools, and data‑aware policy layers from Bltz AI, WideField, Bonfy and Entro, but the piece stresses plumbing (runtime tracing, session auditing) is still unfinished. ( )
Microsoft said more than 80% of Fortune 500 companies now run active AI agents in daily workflows in its Cyber Pulse blog post published February 10, 2026. (microsoft.com) Microsoft has packaged an Agent 365 control plane into a new $99-per-user M365 E7 tier announced March 9, 2026 to give enterprises centralized governance and monitoring for Copilot and agent workloads. (uctoday.com) Bltz AI, founded by former CrowdStrike leaders, launched publicly on March 19, 2026 as an "agentic defensive" platform aimed at securing autonomous AI agents. (prnewswire.com) Bltz bills the product as a self‑healing layer that continuously discovers, tests and automatically remediates vulnerabilities across LLM apps and agentic workflows in real time. (tmcnet.com) WideField closed a Series A round with participation from Cisco Investments and added John Hurley (Optiv CRO) to its board in a March 19, 2026 announcement. (financialcontent.com) As part of that funding-led push, WideField extended its identity-security platform to monitor autonomous AI agents across identities “at rest, in motion, and in use.” (cxodigitalpulse.com) Bonfy published Adaptive Content Security (ACS) 2.0 on March 19, 2026, adding agent‑aware guardrails, an MCP server interface, full Google Workspace parity and SOC 2 Type II certification. (prweb.com) Bonfy’s release also lists integrations with Microsoft Copilot, OpenAI, Anthropic and SIEMs to enable real‑time content protection and blocking for agent-driven flows. (letsdatascience.com) Entro released Agentic Governance & Administration (AGA) on March 18, 2026 to discover AI connections, map non‑human identities and enforce policy at the access‑plane level. (entro.security) Entro frames AGA as an extension of traditional IGA playbooks to answer “who connected what, to which systems, with what permissions” as agent access scales. (markets.businessinsider.com) CrowdStrike’s 2026 Global Threat Report shows AI‑enabled attacks climbed 89% and median breakout time dropped to 29 minutes, underscoring accelerated adversary use of AI across the enterprise attack surface. (crowdstrike.com)
